Case Study 1: Leak on a Lateral

In Case study 1, leak on a lateral connection between a city water main and a private house was determined using acoustic emission method. Test was conducted in the summer of 2015 (on May 22, 2015). Figure 3-3 shows the location of water main and the lateral. As seen in the figure, apparently 18 laterals are connected to the water main between the fire hydrants to supply water to the houses. The lateral with a potential leak is connected to the water main at a distance of 17.8 m from fire hydrant 1 and 83 m from fire hydrant 2 along the length of the water main (from city utility database. A schematic is shown in Figure 3-3). Information of the water main and the lateral is provided in

30

Table 2. Both of the water main and lateral were buried at the depth of 2.43 m to 3 m below ground surface and were backfilled with sandy crushed rock and gravels.

Figure 3-3 Site for leak detection on a lateral

Table 3-2 Information of water main and lateral

Water Main Lateral

Material Ductile Iron Copper

Nominal Diameter 152 mm 19 mm

Wall thickness 9.5 mm 1.65 mm

As a general practice of utilizing Acoustic Emission method, acoustic signals at two fire hydrants are first measured to identify the presence of leaks between the fire hydrants.

31

Figure 4 shows the installation of acoustic sensors on the fire hydrants. As mentioned earlier, acoustic sensors are connected with Wi-Fi modules to send data to a data acquisition system (DAQ) which is connected with a personal computer (laptop computer). A headphone was also connected to the module to hear the noises.

Figure 3-4 Sensors on Fire Hydrant

Acoustic signal recorded in the fire hydrant 1 (sensor 1) and 2 (sensor 2) are shown in Figure 3-5. The signals were recorded at the sampling rate of 11025 data per second. Response spectrums corresponding to the recorded signals are obtained applying Fast Fourier Transform. The resulting frequency spectrum from Fast Fourier Transform is shown in Figure 3-6. Figure 3-6(a) presents the frequency spectrum obtained from the commercial system and Figure 3-6(b) presents the results obtained from MATLAB analysis. The frequency spectrums from the MATLAB analysis and the commercial

32

software are very similar in the figure. Peaks/spikes in the frequency spectrum are observed at almost the same frequencies in both cases (e.g., at 100 Hz, 250 Hz, 300 Hz, 400 Hz and 550 Hz for sensor 2).

33

34

(a) From commercial software

(b) From MATLAB analysis

Figure 3-6 Frequency spectrum of recorded signals

The magnitude-squared coherence between two signals obtained from MATLAB analysis and the coherence from the commercial leak correlator software are included in Figure 3- 7. The magnitude of coherence is around 0.95 in Figure 3-7(a) and (b) within the frequency band of about 500 to 1100 Hz. This clearly indicates that the noises within the frequency band of 500 to 1150 Hz in the two sensors are from the same source, which is

35

potentially a leak. The results obtained from MATLAB analysis are found to correspond to those given by the commercial leak correlator software.

(a) From commercial software

(b) From MATLAB analysis

Figure 3-7 Coherence between two recorded signals

The cross-correlation function in the commercial leak correlation system is provided directly against the distance from the sensor (s). The distance corresponding to the

36

highest magnitude of the cross-correlation is taken as the location of the potential leak. Using MATLAB, the cross-correlation function was calculated against the time lag. The time lag corresponding to the highest magnitude of the cross-correlation is then obtained. This time lag was used to calculate the distance of the noise source from the sensors using Eq. (3-2). Figure 3-8 (a) and (b) shows the cross-correlation functions with time lag from MATLAB and with the distance from a sensor (Sensor 2) obtained for the commercial leak correlator, respectively. In Figure 3-8(a), the magnitude of cross-correlation is the maximum at the time lag of 0.0506 sec. With a wave propagation velocity of 1290 m/s (velocity used in commercial leak finder), the distance to the leak is calculated to be 17.7 m from sensor 1, which is almost the same as the distance obtained from the commercial system (i.e., 17.8 m from sensor 1 in Figure 3-8b). In Figure 3-8(b), the highest magnitude of the cross-correlation is at this distance (i.e., 17.8 m from sensor1 and 82.9 m from Sensor 2).

37

(b) Cross-correlation function with distance from Sensor 2 Figure 3-8 Leak location determination from correlation function

However, since a lateral is connected to the water main at this location (82.9 m from Sensor 2), there was a possibility that the noise source is located on the lateral, which might have propagated into the water main at the intersection. To investigate this further, acoustic emission testing was carried out with one of the sensors (sensor 1) on a fire hydrant (Fire Hydrant 1) and the other (sensor 2) on a curb stop valve on the lateral (see Figure 3-3). Figure 3-9 shows installation of the acoustic sensor on a key-rod connected to the curb stop valve.

38

Figure 3-9 Sensor placed on the key-rod at curb stop valve box

Figures 3-10 and 3-11 show the coherence and cross-correlation of the acoustic signals for this case with one sensor on Fire Hydrant1 and the other on the curb stop valve in the private property. The MATLAB calculations are very similar to the results obtained from the commercial system in Figure 3-10. The coherence magnitudes in the figure are higher within the frequency band of 500 Hz to 1250 Hz and range from 0.5 to 0.7. These coherence magnitudes are somewhat lower than those observed over the similar frequency band (500 to 1250 Hz) when both sensors were on the water main (as discussed above). The lower coherence values in this case are attributed to the burial condition of

39

the lateral, which is different from the water mains, as well the wave propagation through an intersection and different pipe materials.

Figure 3-10 Coherence of signals with a sensor on water main and the other on lateral

However, the coherence values were much higher within the frequency band of 500 Hz to 1250 Hz than the other frequencies in Figure 3-10, indicating a common source of noise within the frequency band.

40

The cross-correlation in Figure 3-11 (a) shows noise source to be located right at curb stop valve (location of sensor 2). This means that the noise source is either located right at the curb stop or between the curb stop and the gate valve in the private house. Similar conclusion can be drawn from the time lag corresponding to the maximum magnitude of cross-correlation in Figure 3-11(b) where the time lag appears to be negative.

41

Another set of acoustic emission testing was then carried out with one sensor at the curb stop valve (sensor 2) and the other on the gate valve at the private house (sensor 1). Figure 3-12 and 3-13 shows coherence and cross-correlation of the acoustic signals. In Figure 3-12, coherence magnitudes of 0.4-0.6 over the 500-900 Hz of frequency band are higher than the coherence magnitudes in other frequencies, indicating again a common source of noise in this frequency band. The magnitude of cross-correlation function is the highest near the curb stop and at the time lag of close to or less than zero in Figure 3- 13. Thus, the location of the noise was expected to be near the curb stop. A leak at this location was later confirmed through excavation.

42

43

Figure 3-13 Leak location determination between curb stop and gate valve